Issues

• offload_state_dict=True leaves CPU model weights on meta after loading

  huggingface/diffusers#15018 · Filter by author Shenrui-Ma opened on Oct 12, 2026

• gfx1201 (RDNA4) SageAttention attention backend: aiter package or Hub kernel?

  huggingface/diffusers#15014 · Filter by author Shan2L opened on Oct 12, 2026
  3

• VaeImageProcessor.denormalize() raises AttributeError for NumPy inputs

  bug
  huggingface/diffusers#15011 · Filter by author SrijanRoy123-github opened on Oct 12, 2026
  1

• [modular] SequentialPipelineBlocks gets required wrong in both inputs and required_inputs

  huggingface/diffusers#15009 · Filter by author yiyixuxu opened on Oct 11, 2026
  1

  

• [Z-Image] joint_attention_kwargs={"scale": ...} is silently ignored when loading LoRA weights

  bug
  huggingface/diffusers#15007 · Filter by author Sergasgr opened on Oct 11, 2026 · Release 0.42.0
  4

• WanVideoToVideoPipeline: "No cache context is set" when caching is enabled

  bug
  huggingface/diffusers#15006 · Filter by author fusheng-ji opened on Oct 11, 2026

• Add TripoSplat

  huggingface/diffusers#15005 · Filter by author hlky opened on Oct 11, 2026
  1

• KDPM2DiscreteScheduler produces NaN on final step due to zero-sigma interpolation

  bug
  huggingface/diffusers#15002 · Filter by author riparanta opened on Oct 10, 2026
  1

• load_video() leaks temporary files when remote video decoding fails

  huggingface/diffusers#14994 · Filter by author SrijanRoy123-github opened on Oct 9, 2026
  2

• WanCausalConv3d copies every activation with F.pad. Let the conv do the spatial padding.

  huggingface/diffusers#14993 · Filter by author will-rice opened on Oct 8, 2026
  1

• Add an optional Transformer Engine FP8 training example for FLUX

  huggingface/diffusers#14992 · Filter by author sbhavani opened on Oct 8, 2026

• LTX-2.5 SDR-To-HDR IC-LoRA: how should it be structured?

  feature-requestlorapipelines
  huggingface/diffusers#14981 · Filter by author christopher5106 opened on Oct 7, 2026
  4

• LTX-2: decouple the RoPE time base from the playback frame rate (Slow-Motion-Control LoRA)

  feature-requestloramodular-pipelinespipelines
  huggingface/diffusers#14980 · Filter by author christopher5106 opened on Oct 7, 2026

• Per-module alphas of kohya text encoder LoRAs collapse to the most common alpha (alpha_pattern keys stop at the parent module)

  buglorapipelines
  huggingface/diffusers#14970 · Filter by author andre-menutole opened on Oct 6, 2026 · Release 0.42.0
  7

• PNDMScheduler.set_timesteps crashes with NumPy broadcasting error for fewer than 4 inference steps

  bugschedulers
  huggingface/diffusers#14953 · Filter by author SrijanRoy123-github opened on Oct 6, 2026
  1

• ModelMixin.from_pretrained on a sharded checkpoint fails under HF_HUB_OFFLINE=1 even when fully cached (_get_checkpoint_shard_files still calls model_info)

  bugmodels
  huggingface/diffusers#14952 · Filter by author quinncheong opened on Oct 5, 2026
  1

• Propose translation to romanian for the diffusers hugging face course

  feature-request
  huggingface/diffusers#14939 · Filter by author pop123-ux opened on Oct 4, 2026

• Qwen-Image LoRAs in lora_A/lora_B format lose their .alpha and load at the wrong strength

  buglorapipelines
  huggingface/diffusers#14938 · Filter by author dkackman opened on Oct 4, 2026

• Qwen-Image-2.1 LoRAs with a fused img_mlp.gate_up load partially without an error

  bugloramodelsneeds-env-infopipelines
  huggingface/diffusers#14937 · Filter by author dkackman opened on Oct 4, 2026

• QwenImage21: model.to(bf16) (as in train_dreambooth_lora_qwenimage21.py) rounds the timestep-embedding freqs buffer, so LoRA training sees a different embedding than inference

  bugmodelsneeds-env-info
  huggingface/diffusers#14935 · Filter by author tritsystem opened on Oct 4, 2026

• StableAudio3DiTModel: loading with torch_dtype=bf16/fp16 collapses RoPE positions (arange built in inv_freq dtype)

  bugmodels
  huggingface/diffusers#14934 · Filter by author tritsystem opened on Oct 4, 2026

• from_single_file() support for AnimaTextConditioner

  feature-requestmodelssingle-file
  huggingface/diffusers#14930 · Filter by author mingyi456 opened on Oct 3, 2026
  2

• CosmosTransformer3DModel.from_single_file() does not work with the model.diffusion_model. prefix

  bugmodelssingle-file
  huggingface/diffusers#14929 · Filter by author mingyi456 opened on Oct 3, 2026

• Qwen-Image-2.1 VAE tiling: RGB/alpha seams and a full-context decode workaround on ROCm

  bugmodelsneeds-code-example
  huggingface/diffusers#14928 · Filter by author nabe2030 opened on Oct 3, 2026

• Windows: from_single_file original_config and fp16_safetensors crash with UnicodeDecodeError under non-UTF-8 locales

  bugmodelsneeds-code-exampleneeds-env-infosingle-file
  huggingface/diffusers#14927 · Filter by author dltsum opened on Oct 3, 2026